TBR1 Directly Represses Fezf2 to Control the Laminar Origin and Development of the Corticospinal Tract

TBR1 directly represses Fezf2 to control the laminar origin and development of the corticospinal tract

Wenqi Han1, Kenneth Y. Kwan1, Sungbo Shim1, Mandy M. S. Lam, Yurae Shin, Xuming Xu, Ying Zhu, Mingfeng Li, and Nenad Šestan2

Department of Neurobiology, Kavli Institute of Neuroscience, Yale University School of Medicine, New Haven, CT 06510

Edited by Edward G. Jones, University of California, Davis, CA, and approved January 7, 2011 (received for review November 9, 2010) The corticospinal (CS) tract is involved in controlling discrete concomitant with a decrease in Sox5 expression. Those findings, voluntary skilled movements in mammals. The CS tract arises along with the observation that Tbr1 null mutants exhibit axon exclusively from layer (L) 5 projection neurons of the cerebral projection defects with some similarities to those of the Sox5 null cortex, and its formation requires L5 activity of Fezf2 (Fezl, phenotype (17, 18), led to the hypothesis that Tbr1 regulates Zfp312). How this L5-specific pattern of Fezf2 expression and CS Fezf2 expression indirectly via Sox5. axonal connectivity is established with such remarkable fidelity In the present study, we tested the alternative hypothesis that had remained elusive. Here we show that the transcription factor Tbr1 directly represses Fezf2 transcription and thereby controls TBR1 directly binds the Fezf2 locus and represses its activity in L6 the laminar origin and development of the CS tract. We found corticothalamic projection neurons to restrict the origin of the CS that TBR1 binds directly to a conserved regulatory element near tract to L5. In Tbr1 null mutants, CS axons ectopically originate the Fezf2 gene and selectively represses its activity in L6 cortico- from L6 neurons in a Fezf2-dependent manner. Consistently, mis- thalamic projection neurons. Moreover, in Tbr1 null mutants, CS expression of Tbr1 in L5 CS neurons suppresses Fezf2 expression axons originate ectopically from L6 neurons in a Fezf2-dependent and effectively abolishes the CS tract. Taken together, our findings manner. Consistent with this, TBR1 was absent from Fezf2- show that TBR1 is a direct transcriptional repressor of Fezf2 and expressing L5 CS tract neurons during development, and the a negative regulator of CS tract formation that restricts the lami- forced expression of Tbr1 in these neurons suppressed Fezf2 and nar origin of CS axons specifically to L5. blocked CS tract formation. Taken together, our findings show that TBR1 is a direct transcriptional repressor of Fezf2 that blocks neocortex | pyramidal neuron | axon guidance | transcriptional repression the formation of the CS tract from L6, thereby restricting the origin of CS axons specifically to L5. patial specificity of axonal connections is one of the most Simportant prerequisites for normal development (1–3). In Results mammals, this is especially crucial for axons of the corticospinal TBR1 Binds Directly to a Consensus Site Near Fezf2. Using micro- (CS) system (4–7). Development of the CS tract is an intricate arrays to analyze changes in the Tbr1 null cortical transcriptome, process that involves the molecular specification of CS neurons Bedogni et al. (17) found a significant increase in Fezf2 expression, and axon pathfinding. All long-range subcortical axons projecting and suggested that this up-regulation is a result of decreased Sox5 to the brainstem and spinal cord, including those that form the expression. Our independent analyses using mRNA-Seq (Fig. CS tract, originate exclusively from layer (L) 5 projection (py- 1A), quantitative RT-PCR (qRT-PCR; Fig. 2A), and in situ hy- B fi fi ramidal) neurons of the cerebral cortex (8–11). Projection neu- bridization (ISH; Fig. 2 ) con rmed a signi cant up-regulation of Fezf2 Tbr1−/− rons in other cortical layers give rise to axons that project within in neocortex. However, analysis by immunostaining and immunoblotting revealed that SOX5 protein levels were rel- the cortex (L2–4) or to the thalamus (L6). How this highly − − atively unaltered in late embryonic and neonatal Tbr1 / cortex conserved laminar pattern of projections is formed with such Fezf2 perfect accuracy remains elusive. (Fig. S1). This led us to the hypothesis that TBR1 regulates transcription via direct binding to regulatory sequences near Previous work revealed that the transcription factor FEZF2 Fezf2. To identify genome-wide TBR1 binding sites in an unbiased (FEZL, ZFP312) is highly enriched in L5 CS neurons and is – and hypothesis-independent manner, we analyzed TBR1-immu- critical to the development of the CS tract (12 14). Inactivation noprecipitated chromatin using deep sequencing (ChIP-Seq; Fig. of Fezf2 disrupts formation of the CS tract (12–14), whereas B Fezf2 1 ). We tested several available anti-TBR1 antibodies and found misexpression of in upper layer projection neurons induces that none was suitable for immunoprecipitating chromatin of fi ectopic subcortical projections (13). These ndings indicate that sufficient quality for ChIP-Seq. Thus, we generated a V5-TBR1 Fezf2 transcription is tightly regulated during development, and fusion construct and expressed it in N2A cells. V5-TBR1 was that the integrity of normal Fezf2 expression is critical to proper immunoprecipitated using an anti-V5 antibody. DNA-Seq was development of the CS tract. Interestingly, Fezf2 is transiently performed on the Illumina GAIIx platform, and data were ana- expressed in L6 neurons during early embryonic development, lyzed as described in SI Materials and Methods. A total of 34.5 where its transcription is directly repressed by SOX5, thereby million reads of 36 bp were generated for V5-precipitated chro- establishing a high-in-L5, low-in-L6 postnatal pattern (15, 16). matin and 37 million reads were generated for control input DNA, Paradoxically, in Sox5 null mutants, the number of axon projections reaching the brainstem and spinal cord is severely reduced despite increased cortical Fezf2 expression (15). This Author contributions: W.H., K.Y.K., and N.Š. designed research; W.H., K.Y.K., S.S., M.M.S.L., suggests that Sox5 is required for the formation of these con- and Y.S. performed research; W.H., K.Y.K., X.X., Y.Z., M.L., and N.Š. analyzed data; and nections independent of its regulation of Fezf2. Furthermore, W.H., K.Y.K., and N.Š. wrote the paper. SOX5 is normally expressed in L5 Fezf2-expressing CS neurons The authors declare no conflict of interest. (15). Therefore, the down-regulation of Fezf2 in L6 neurons, and This article is a PNAS Direct Submission. hence the establishment of Fezf2’s L5 pattern, likely requires Freely available online through the PNAS open access option. additional molecules. Recently, Bedogni et al. (17) showed that 1W.H., K.Y.K., and S.S. contributed equally to this work. TBR1 exerts positive and negative control over a number of 2To whom correspondence should be addressed. E-mail: [email protected]. genetic markers of regional (areal) and laminar identity. Nota- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. NEUROSCIENCE bly, Tbr1 null mutants exhibit an increase in Fezf2 expression 1073/pnas.1016723108/-/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.1016723108 PNAS | February 15, 2011 | vol. 108 | no. 7 | 3041–3046 Fig. 1. TBR1 regulates Fezf2 transcription via direct binding and repression of regulatory elements near Fezf2.(A) RNA-Seq of mRNA extracted from − − − − neocortices of P0 Tbr1 / and Tbr1+/+ littermates (n = 2 per genotype). Significantly more RNA-Seq reads mapped to Fezf2 exons (gray) in Tbr1 / (green; 727 reads) than in Tbr1+/+ (red, 496 reads; adjusted P value = 1.49E-9). (B) N2a cells were transfected with V5-TBR1, and chromatin was immunoprecipated using an anti-V5 antibody. ChIP-Seq revealed enrichment of mapped reads (blue) compared with input DNA in a region 4.5 kb downstream of the Fezf2 TSS. (C) Schematic of the mouse Fezf2 locus (gray) and mammalian conservation (dark blue) showing the location of the TBR1 consensus binding site (red) defined by ChIP-Seq (detailed in Fig. S2). (D) ChIP–qRT-PCR assays in neurons cultured from E13.5 neocortex transfected with V5-TBR1. DNA precipitated with control rabbit IgG or anti-V5 antibody was analyzed by qRT-PCR using primers flanking the TBR1 binding site and control primers. The TBR1 site was significantly enriched in V5-precipitated DNA. (E) N2A cells were transfected with luciferase plasmids containing the TBR1 binding site. Luciferase assay showed that cotransfection of CAG-Tbr1, but not of empty CAG, repressed the activity of this element. *P < 0.02; **P < 0.005 by unpaired two-tailed t test. ns, not significant (P > 0.05).

− of which 82.9% and 85.1%, respectively, uniquely mapped to the neurons, we bred Tbr1+/ mice with Fezf2-Gfp transgenic mice. − − genome. Of immunoenriched regions with a false discovery rate In the postnatal day (P) 0 Tbr1 / neocortex, the number of cutoff of 0.02, approximately 62% were mapped within gene loci, neurons highly expressing Fezf2-Gfp, which did not migrate +/+ and an additional 28% were mapped within 50 kb upstream of normally (18), increased significantly from 21.8% in Tbr1 to −/− transcription start sites (TSSs) or 50 kb downstream of the 33.3% in Tbr1 (P=0.0058) (Fig. 2C). This significant increase − − polyadenylation signal. in the total number of Fezf2-Gfp–expressing neurons in Tbr1 / To identify putative TBR1 consensus binding sequences with neocortex reflects an aberrant induction of high Fezf2 expression potential relevance to TBR1 transcriptional activity in the neo- in neurons that do not normally express Fezf2. cortex, we cross-analyzed our mRNA-Seq and ChIP-Seq data. Because TBR1 is specifically expressed in L6 neurons at P0, We selected the loci of significantly differentially expressed we examined whether L6 neurons aberrantly turned on high − − genes based on mRNA-Seq data to which a significant enrich- Fezf2 expression in the Tbr1 / neocortex. We used two stringent − − ment of TBR1 ChIP-Seq reads mapped proximally. Analysis of criteria to conservatively define L6 neurons in the Tbr1 / neo- these 181 loci (detailed in SI Material and Methods) revealed cortex, in which the demarcation of distinct layers is complicated a binding sequence motif (Fig. S2) with similarities to the T el- by migration defects (ref. 18 and Fig. S4). First, we chose ement in Xenopus (19, 20). Importantly, an enrichment of TBR1 ZFPM2 (FOG2) to molecularly define L6 neurons, because it is ChIP-Seq reads compared with input DNA was mapped to a reliable marker of L6 corticothalamic neurons in the early a highly conserved region ∼4.5 kbp downstream of the Fezf2 postnatal neocortex (ref. 15 and Fig. S5) and, more importantly, − − TSS. This region contained a core consensus binding site con- because its expression level is unchanged in Tbr1 / neocortex served across placental mammalian species (Fig. 1C and Fig. (Fig. S6, mRNA-Seq; P=0.642 and ref. 17). Second, we used S2C). To confirm this binding in the neocortex, we performed a stringent birth-dating strategy to distinguish L6 neurons from ChIP using cultured embryonic day (E) 13.5 neocortical neurons L5 neurons. In the sequential generation of cortical neurons, the transfected with V5-TBR1. Analysis of ChIPed DNA by qRT- same progenitor cells that give rise to L6 neurons subsequently PCR revealed a significant enrichment of this binding site in give rise to L5 neurons. Therefore, positive thymidine analog DNA precipitated by V5 antibody compared with control IgG (a incorporation of L6 neurons will inevitably label a large number 6.65 ± 0.37-fold increase; P=0.00128) (Fig. 1D), confirming of L5 neurons. To circumvent this, we used exclusion of the that TBR1 binds this region in neocortical neurons. To assess thymidine analog CldU to distinguish L6 and L5 neurons. CldU − whether this binding affects transcription, we cloned the binding was administered to timed-pregnant Tbr1+/ females daily site downstream of the luciferase reporter gene and assayed starting at E12.5, when the first L5 neurons are born, and ending transfected N2a cells. Forced expression of Tbr1 moderately but at E14.5, when the last L5 neurons are generated. Using this − significantly repressed the activity of this binding site (a decrease narrow definition of L6 neurons (ZFPM2+, CldU ), we observed of 31.9% ± 8.6%; P=0.0105) (Fig. 1E). Collectively, these asignificant increase in the number of L6 neurons expressing − − findings suggest that TBR1 negatively regulates Fezf2 transcrip- Fezf2-Gfp (18.5% in Tbr1+/+ and 85.9% in Tbr1 / ; P= tion through direct binding to a conserved consensus element. 0.000031) (Fig. 2D). Thus, in the absence of Tbr1, neurons born at the correct time and with the molecular properties of L6 L6 Neurons Aberrantly Retain High Fezf2 Expression in Tbr1 Null neurons ectopically expressed high levels of Fezf2. Mutants. The global increase in Fezf2 expression in the early − − postnatal Tbr1 / neocortex found in the present study and Ectopic L6 Projections to the Brainstem Are Revealed by Retrograde a previous study (17) could result from either up-regulation of Tracing in Tbr1 Null Mutants. Subcortical axon defects have been − − Fezf2 in neurons that normally express Fezf2 or ectopic expres- reported in Tbr1 / brains (18). Long-range subcerebral axons sion of Fezf2 in neurons that do not normally express Fezf2. and the layer-specific pattern of cortical connectivity have not Consistent with the latter possibility, analysis of TBR1 expres- been specifically examined, however. To test whether aberrantly sion during cortical development revealed that TBR1 and Fezf2 high expression of Fezf2 in L6 neurons leads to the formation of − established complementary laminar expression patterns during ectopic projections to the brainstem, we bred Tbr1+/ mice with late embryogenesis (Fig. S3A). To directly assess how TBR1 Golli-Gfp transgenic mice, which express GFP predominantly in alters Fezf2 expression at the level of resolution of individual L6 and SP neurons and their axons (21), and with Fezf2-Gfp

3042 | www.pnas.org/cgi/doi/10.1073/pnas.1016723108 Han et al. Fig. 3. Premature overgrowth and superabundance of subcortical axons − − in the brainstem of Tbr1 / mice. (A) In E14.5 Tbr1+/+, Golli-Gfp+ axons originating mainly from L6 neurons extended to the upper portion of the striatum (Str; above the dashed line). In Tbr1−/−, numerous Golli-Gfp+ axons extended to the deeper portion of the striatum (below the dashed − − line) toward the internal capsule. (B)InP0Tbr1 / , Golli-Gfp+ axons were not present in the dorsal thalamus (Th), indicating that the normal corticothalamic tract (solid arrowhead) had not formed. Ectopic axons (arrow) were misdirected toward other subcortical targets, including the hypothalamus (Hyp) and the midbrain. (C) In E16.5 Tbr1+/+ brain, Fezf2-Gfp+ axons (solid arrowheads) did not cross the forebrain/midbrain boundary − − (dashed line). In Tbr1 / , numerous Fezf2-Gfp+ axons (open arrowheads) extended subcerebrally to invade the midbrain (MB). (D) In E16.5 Tbr1+/+ brain, only a few Fezf2-Gfp+ axons innervated the CPs (dashed lines). − − Dramatically more Fezf2-Gfp+ axons entered the CPs in Tbr1 / . (Scale bars: 100 μm.)

aberrantly project axons to the brainstem due to increased Fezf2 Fig. 2. High Fezf2 expression is aberrantly retained in neonatal L6 neurons − − Golli-Gfp in Tbr1 / neocortex. (A) qRT-PCR of P0 neocortical mRNA. Normalized to expression. Although the transgene is a useful marker of Gapdh Fezf2 ± Tbr1−/− L6 neurons, a small number of neurons near the L5/L6 border of , expression levels increased by 1.77 0.08-fold in Golli-Gfp compared with Tbr1+/+.(B) Fezf2 ISH of P0 brains showed increased Fezf2 the cingulate cortex express and project axons to the −/− Golli-Gfp expression in Tbr1 neocortex. (Scale bar: 500 μm.) (C) P0 neocortices of brainstem (15). Therefore, the presence of exuberant −/− Tbr1+/+ and Tbr1−/− littermates transgenic for Fezf2-Gfp. Fezf2-Gfp– axons in the Tbr1 brainstem does not unequivocally indicate expressing neurons were dispersed in the Tbr1−/−, consistent with previously that the axons originated from L6 neurons. Thus, we performed reported defects in neuronal migration (18). The proportion of neurons retrograde axonal tracing by injecting rhodamine-conjugated la- +/+ −/− expressing Fezf2-Gfp significantly increased from 21.8% in the Tbr1+/+ to tex microspheres (LMS) into the pons of Tbr1 and Tbr1 − − 33.3% in the Tbr1 / (1.53 ± 0.27-fold). (Scale bar: 50 μm.) (D) L6 neurons newborn mice. Analysis was carried out after 18–24 h of survival to (arrowhead) were stringently defined by expression of the L6 cortico- allow for transport of the tracer, a time frame that is limited by the − − thalamic neuron marker ZFPM2 (red) and a lack of incorporation of CldU perinatal lethality of the Tbr1 / (22). In addition to the highly (blue), multiple doses of which were injected daily from E12.5 to E14.5, when stringent criteria described above for identifying L6 neurons using L5 neurons were generated. The number of L6 neurons expressing Fezf2-Gfp ZFPM2 immunostaining and exclusion of E12.5–E14.5 CldU in- fi Tbr1+/+ Tbr1−/− increased signi cantly from 18.5% in the to 85.9% in the corporation, we further narrowed our definition using Golli-Gfp, (4.64 ± 0.20-fold). (Scale bar: 10 μm.) *P < 0.01; **P < 0.00005 by unpaired t which is mostly restricted to L6 neurons in the somatosensory two-tailed test. cortical areas that we analyzed. In the Tbr1+/+, projection neurons labeled retrogradely from transgenic mice. In E14.5 normal embryos, Golli-Gfp axons en the pons were restricted to L5, and most incorporated CldU and expressed neither Golli-Gfp nor ZFPM2 (Fig. 4B, open arrow- route to the thalamus entered the upper portion of the striatum, −/− heads). A limited number of labeled neurons were positive for but not the lower portion (Fig. 3A). In E14.5 Tbr1 brains, Golli-Gfp Golli-Gfp or lacked CldU, but none expressed ZFPM2. Most axons exhibited premature overgrowth, ectopically importantly, as defined by the highly conservative triple criteria − invading the lower portion of the striatum and the primitive in- (Golli-Gfp+, ZFPM2+, and CldU ), no L6 neurons were labeled ternal capsule. At P0, numerous Golli-Gfp axons reached the Tbr1+/+ n Tbr1+/+ B by the retrograde tracer in any of the traced brains ( = dorsal thalamus in (Fig. 3 , solid arrowhead). In the P0 8). In contrast, defined by the same stringent criteria, a consider- Tbr1−/− Golli-Gfp − − brains, none of the axons entered the dorsal able number of L6 neurons in the Tbr1 / somatosensory cortex thalamus, but many were aberrantly directed toward other sub- were clearly labeled by the retrograde tracer, which encircled cortical targets, including the hypothalamus (Fig. 3B, arrow) and neuronal nuclei and entered apical dendrites (Fig. 4C, solid the midbrain. Similarly, exuberant Fezf-Gfp axons in E16.5 and arrowheads; n = 4 brains). Similar to reported retrograde tracing −/− − − P0 Tbr1 brains (Fig. 3C, open arrowheads and Fig. S7), ec- from the CPs of the Tbr1 / (18), retrograde tracing from the pons topically innervated the cerebral peduncles (CPs; Fig. 3D, bro- was absent from the most medial cortical areas. Thus, the so- ken line) and brainstem, indicating that the prematurely matosensory cortex was consistently used for quantification, which outgrown and misguided L6 axons ectopically expressed high revealed a significant increase in the number of labeled L6 neu- − − levels of Fezf2-Gfp. rons in the Tbr1 / for each of the three criteria used (Golli-Gfp+, − Normal L6 projection neurons project axons only as far as the P=0.000009; ZFPM2+, P=0.000007; CldU , P=0.000049) thalamus, never reaching the brainstem (8). Our data are consis- (Fig. 4D). Notably, the numbers of retrogradely labeled Golli- NEUROSCIENCE − − − − − tent with the intriguing possibility that L6 neurons in the Tbr1 / Gfp and LSM+ neurons were unaltered in the Tbr1 / somato-

Han et al. PNAS | February 15, 2011 | vol. 108 | no. 7 | 3043 − − Fig. 4. Stringently defined L6 neurons aberrantly formed subcerebral/CS projections in Tbr1 / .(A) Schematic of the experimental design. CldU was ad- − ministered daily between E12.5 and E14.5 to label L5 neurons in timed-pregnant Tbr1+/ females transgenic for Golli-Gfp. Retrograde tracer (LMS) was injected into the pons of live neonates. L6 neurons were narrowly defined by lack of CldU incorporation and coexpression of Golli-Gfp and ZFPM2. (B and C) Confocal images of the somatosensory cortex were obtained at an optical thickness of 3 μm. (B)InTbr1+/+, all retrogradely labeled neurons were positioned in − − L5, and most were Golli-Gfp and ZFPM2 and incorporated CldU (open arrowhead). No neurons meeting our stringent definition of L6 neurons were − − retrogradely traced in any of the Tbr1+/+ brains analyzed (n = 8). (Scale bar: 50 μm.) (C)InTbr1 / , a significant number of L6 neurons meeting our criteria were labeled retrogradely (solid arrowhead; n = 3 brains). Ectopic CS projections of these L6 neurons indicate a projectional fate switch. (Scale bar: 50 μm). (D and −/− +/+ E). Quantitative analysis revealed that the number of traced L6 neurons increased significantly in Tbr1 compared with Tbr1 for each of the criteria used. However, no difference in the number of traced Golli-GFP− L5 neurons was detected. (F) Analysis of nucleus diameter. In Tbr1+/+ neocortex, L5 CS neurons (LMS+) were significantly larger in nuclear diameter compared with L6 corticothalamic neurons (CTh; ZFPM2+ and Golli-Gfp+). In Tbr1−/−, nuclei of L6 CS neurons (LMS+, ZFPM2+, and Golli-Gfp+) were indistinguishable in size from Tbr1+/+ L6 CTh neurons but significantly smaller than Tbr1+/+ L5 CS neurons. *P < 0.00005 by unpaired two-tailed t test. ns, not significant (P > 0.05). sensory cortex (Fig. 4E), suggesting that L5 projections to the pons decreased dramatically, to 10.8% ± 4.2% (P=0.000003) (Fig. 5 were mostly normal. We also used nuclear size to confirm the A and B). This indicates that Tbr1 misexpression in L5 neurons is identity of the CS neurons projecting to the brainstem and spinal sufficient to suppress Fezf2 expression. − − cord. Normal L5 CS neurons (LMS+, ZFPM2 , and Golli-Gfp ) To investigate whether this repressed Fezf2 expression func- have larger nuclei compared with L6 corticothalamic neurons tionally affects the axonal projections of L5 CS neurons, we − (LMS ,ZFPM2+,andGolli-Gfp+; P=2.25E-17) (Fig. 4F). performed in utero electroporation to transfect L5 neurons with Analysis of retrogradely traced L6 neurons (LMS+, ZFPM2+, and CAG-Tbr1 or empty CAG plasmid DNA, together with CAG- + − − Golli-Gfp ) in the Tbr1 / revealed a nuclear diameter in- Gfp, because the higher solubility of GFP better facilitates the distinguishable from that of normal L6 corticothalamic neurons analysis of long-distance CS axons. At E13.5, neurons electro- (P=0.168), but significantly smaller than that of L5 CS neurons porated with empty CAG plasmid projected corticocortical − − (P=2.11E-13). These data indicate that in the Tbr1 / neocortex, axons into the corpus callosum (Fig. 5D) and projected numer- L6 neurons, as stringently defined by marker expression, birth- ous CS axons into the ventral pons and medullary pyramids (Fig. dating, and nuclear size, project axons ectopically to the brainstem 5 E and F). In contrast, GFP+ axons originating from L5 neurons instead of to the thalamus and express high levels of Fezf2, key misexpressing Tbr1 were almost completely absent from the pons features of normal L5 neurons. and pyramids. This disruption of axons was specific to the CS tract, however. Tbr1-misexpressing neurons projected abundant Ectopic Tbr1 Expression in L5 Neurons Effectively Abolishes Fezf2 GFP+ axons into the corpus callosum (Fig. 5D). Together, these Expression and CS Tract Formation. TBR1 is normally excluded findings demonstrate that the misexpression of Tbr1 is sufficient from L5 Fezf2-expressing CS neurons during embryonic and to suppress Fezf2 expression in L5 neurons and effectively postnatal development (Fig. S3 A and B). To investigate whether abolish formation of the CS tract. forced misexpression of Tbr1 in L5 neurons is able to down- regulate Fezf2 and suppress CS projections, we performed in Induction of Ectopic L6 CS Axons in the Tbr1 Null Mutants Is Depen- utero electroporation at E13.5 to transfect L5 neurons with dent on Fezf2. To examine whether the induction of ectopic axons −/− CAG-Tbr1 or control empty CAG plasmid DNA, together with originating from L6 neurons in the Tbr1 cortex depends on CAG-Rfp to label neurons generated by electroporated pro- up-regulation of Fezf2, we engineered a conditional Fezf2 allele genitor cells. To analyze Fezf2 expression in transfected cells, we (Fig. S8 A and B) for the generation of a cortex-specific Tbr1; flox/flox carried out these experiments in Fezf2-Gfp transgenic mice. The Fezf2 double-null mutant. Fezf2 mice were crossed with vast majority of L5 RFP+ neurons electroporated with empty Emx1-Cre PAC transgenic mice (23) to specifically inactivate CAG plasmid expressed high levels of Fezf2-Gfp (85.3% ± 4.3%) Fezf2 in the cortex (Fig. S8 A and B). To confirm that our cortex- (Fig. 5 A and B). Remarkably, when Tbr1 was misexpressed, the specific conditional Fezf2 null mutant was phenotypically similar − − number of L5 RFP+ neurons expressing high levels of Fezf2-Gfp to germ-line Fezf2 / mice (12, 14), we crossed Emx1-Cre PAC;

3044 | www.pnas.org/cgi/doi/10.1073/pnas.1016723108 Han et al. lator of the CS tract that restricts the tract’s laminar origin specifically to L5. Our data confirm and extend those of previous studies showing the role of Tbr1 in the formation of subcortical projections (17, 18). We show here that TBR1 directly represses Fezf2 and TBR1 overexpression in L5 neurons abolishes CS projections. However, given that L5 CS neurons normally extend collateral axon branches to the dorsal thalamus (25), exogenous expression of TBR1 in these neurons cannot be used to address whether TBR1 is sufficient to confer a corticothalamic projection fate. Notably, the upper-layer neurons, which do not project to the thalamus, normally express TBR1 starting at an early postnatal age. Therefore, it is likely that TBR1, although necessary for normal corticothalamic projections (17, 18), is not sufficient to induce this projection fate. Using the Fezf2-Gfp transgene to reveal the global pattern of − − corticofugal connections, we found that at P0, the Tbr1 / CS tract is largely of normal size compared with controls (Fig. S7). The additional axons projecting to the brainstem from L6 neurons might be expected to increase the size of this tract in the − − Tbr1 / ; however, this increase is likely offset by the role of Tbr1 Fig. 5. TBR1 misexpression in L5 neurons abolished Fezf2 expression and in areal patterning (17, 18), in which it is required by neurons in ablated CS axons. (A and B) CAG-Tbr1 or empty CAG plasmid DNA was de- Rfp Fezf2-Gfp the medial cortex to project subcortical axons. Taken together, livered together with CAG- into E13.5 embryos expressing these findings indicate that TBR1 is a critical determinant of using in utero electroporation. RFP-expressing L5 neurons were examined at P0. Of the L5 neurons electroporated with empty CAG plasmid, 85.3% ± multiple aspects of cortical development, including neuronal Fezf2-Gfp ± Tbr1 migration, laminar and areal identity, and axonal projection. 4.3% expressed . In contrast, only 10.8% 4.2% of the -mis- Fezf2 fi expressing L5 neurons expressed Fezf2-Gfp, a significant reduction com- The dynamic pattern of exempli es how precise temporal pared with empty CAG. P=0.000003. (C–F) CAG-Tbr1 or CAG-empty plasmid and spatial regulation of gene expression has functional con- DNA was coelectroporated with CAG-Gfp into WT E13.5 embryos. Brains sequences in neocortical development and axonal connections. were collected for analysis at P0. (C) The majority of neurons expressing Furthermore, the exclusively postmitotic expression pattern of CAG-Tbr1 migrated normally to settle in all cortical layers, including L5. In all Tbr1 (17, 18) indicates that layer-specific molecular and pro- examined brains that overexpressed Tbr1, ectopic neuropils and cells were jectional identities are not firmly established during neurogenesis, clustered in the white matter (*); however, most axons entered the white but undergo postmitotic refinement to establish their mature matter and innervated the corpus callosum (D). Neurons electroporated with patterns. This fine-tuning of neuronal identity and connectivity empty CAG plasmid projected numerous subcerebral/CS axons into the provides a mechanism by which the precise configuration of ventral pons (empty arrowhead) and medullary pyramids. Misexpression of neural circuits can be modified after neurons integrate into Tbr1 ablated all CS axons. A few GFP+ axons reached the ventral pons (E), but none reached the medullary pyramids (F). This indicates that mis- the cortex. Fezf2-Gfp TBR1 is not the only molecule that postmitotically controls expression of TBR1 in L5 neurons substantially down-regulated Fezf2 and blocked formation of the CS tract. (Scale bars: 20 μminA; 200 μminC–E; the dynamic pattern of expression. Another transcription 50 μminF.) factor, SOX5, contributes to this process as well (15, 16). In contrast to the Tbr1 null mutant, in the Sox5 null mutants projections to the brainstem are nearly completely abolished (15), flox/flox Sox5 Fezf2 mice with CAG-Cat-Gfp transgenic mice (24) to re- indicating that is independently required for the formation Fezf2 veal the pancortical pattern of axonal connectivity. Similar to of CS tract. Moreover, SOX5 is normally expressed in - − − germ-line Fezf2 / , the CS tract was nearly completely absent expressing L5 CS neurons (15). These results and our analysis of C–E TBR1 binding sites by ChIP-Seq and ChIP–qRT-PCR are con- from the brainstem (Fig. S8 ). For analysis of L6 neurons, we Fezf2 bred in the Golli-Gfp transgene to generate cortex-specific Tbr1; sistent with the hypothesis that TBR1 regulation of tran- − − fl/fl Fezf2 double-null mutants (Tbr1 / , Fezf2 , Emx1-Cre, and scription is direct rather than secondary to changes in SOX5 Golli-Gfp). LMS was injected in the pons of Golli-Gfp double- levels, as was previously suggested. Collectively, these data indicate that both SOX5 and TBR1 are required in the regulation null mice and littermate controls at P0. Remarkably, neither L5 Fezf2 nor L6 projection neurons were retrogradely labeled from the of transcription. Whether the two transcription factors physically interact or cooperate in alternative ways in this role pons of the double-null mutants (Fig. 6). Collectively, these data remain unknown, however. indicate that cortical Fezf2 activity is required for the formation All projections to the brainstem, including the CS system, of ectopic L6 projections to the pons, including the CS system originate from L5 neurons in all mammalian species examined axons in Tbr1 null mutants, and that Tbr1 restricts the laminar – fi Fezf2 (4 6). The high degree of conservation indicates that this speci c origin of the CS tract by repressing in L6 neurons in pattern of connectivity is likely critical to the survival and evolu- normal mice. tionary success of mammals. It also suggests that the regulatory Discussion pathways that establish this pattern are well conserved. Consistent with this, the prenatal neocortical expression patterns of Fezf2, We have shown that TBR1 directly binds a conserved consensus Tbr1, and Sox5 are similar in different mammals, including hu- site near the Fezf2 locus and represses Fezf2 expression in L6 Tbr1 mans (15, 26, 27). Therefore, TBR1-mediated down-regulation of neocortical projection neurons. In the absence of , postnatal Fezf2 in L6 is most likely a conserved mechanism for restricting Fezf2 L6 neurons aberrantly express high levels of and ectopi- the laminar origin of the CS tract to L5. cally project to the brainstem. These ectopic L6 projections to the brainstem, including the CS axons, are Fezf2-dependent. Materials and Methods fi Fezf2 Moreover, TBR1 misexpression is suf cient to repress in Animals. All experiments were carried out in accordance with a protocol L5 neurons and block the formation of the CS tract. These approved by Yale University’s Committee on Animal Research. Tbr1 mice

ﬁndings (summarized in Fig. S9) demonstrate that TBR1 is were a generous gift from John Rubenstein (22). Fezf2-Gfp mice were NEUROSCIENCE a direct transcriptional repressor of Fezf2 and a negative regu- obtained from GENSAT (28). Golli-Gfp, Emx1-Cre PAC, and CAG-Cat-Gfp

Han et al. PNAS | February 15, 2011 | vol. 108 | no. 7 | 3045 − − Fig. 6. Aberrant L6 projections to the brainstem in the Tbr1 / require Fezf2. Cortex-specific double deletion of Fezf2 and Tbr1 was achieved using the Emx1- Cre transgene. LMS was injected into the pons of neonatal mice of the indicated genotype. (A)InTbr1+/+ mice, retrogradely labeled neurons were positioned − − in L5 and did not express Golli-Gfp or ZFPM2 (open arrowheads). (B)InTbr1 / mice, many L6 neurons expressing Golli-Gfp and ZFPM2 were retrogradely − − fl/fl labeled by LMS (solid arrowheads). (C)InTbr1 / ;Fezf2 ;Emx1-Cre cortex-specific double mutant, no Golli-Gfp+ or ZFPM2+ L6 neurons were retrogradely labeled. These data indicate a partial rescue of the Tbr1−/− phenotype in the absence of cortical Fezf2 and suggest that the L6 CS projections of Tbr1−/− depend on cortical Fezf2. (Scale bars: 200 μm.) mice were generous gifts from Anthony Campagnoni (21), Takuji Iwasato ACKNOWLEDGMENTS. We thank A. Louvi, D. McCormick, and members (23) and Melissa Colbert (24), respectively. of the N.Š. laboratory for constructive discussions. We especially thank Further experimental details can be found in SI Materials and Methods. Y. Imamura Kawasawa for preparing sequencing libraries and S. Mane and M. Mahajan for performing sequencing. This work was supported by a grant from the US National Institutes of Health (NS 054273). Note Added in Proof. After the approval of this paper for publication, similar W.H. and Y.Z. are supported by predoctoral fellowships from the China findings by McKenna et al. (29) showed that Tbr1 and Fezf2 regulate the Scholarship Council, and Y.S. is supported by a predoctoral fellowship alternate corticofugal identities of projection neurons. from Samsung.

1. Parrish JZ, Emoto K, Kim MD, Jan YN (2007) Mechanisms that regulate establishment, 16. Lai T, et al. (2008) SOX5 controls the sequential generation of distinct corticofugal maintenance, and remodeling of dendritic fields. Annu Rev Neurosci 30:399–423. neuron subtypes. Neuron 57:232–247. 2. Polleux F, Ince-Dunn G, Ghosh A (2007) Transcriptional regulation of vertebrate axon 17. Bedogni F, et al. (2010) Tbr1 regulates regional and laminar identity of postmitotic Nat Rev Neurosci – guidance and synapse formation. 8:331 340. neurons in developing neocortex. Proc Natl Acad Sci USA 107:13129–13134. fi Annu Rev Cell Dev 3. Sanes JR, Yamagata M (2009) Many paths to synaptic speci city. 18. Hevner RF, et al. (2001) Tbr1 regulates differentiation of the preplate and layer 6. Biol – 25:161 195. Neuron 29:353–366. 4. ten Donkelaar HJ, et al. (2004) Development and malformations of the human 19. Kispert A, Hermann BG (1993) The Brachyury gene encodes a novel DNA binding pyramidal tract. J Neurol 251:1429–1442. protein. EMBO J 12:4898–4899. 5. Martin JH (2005) The corticospinal system: From development to motor control. 20. Wang TF, et al. (2004) Identification of Tbr-1/CASK complex target genes in neurons. Neuroscientist 11:161–173. J Neurochem 91:1483–1492. 6. Lemon RN (2008) Descending pathways in motor control. Annu Rev Neurosci 31:195–218. 21. Jacobs EC, et al. (2007) Visualization of corticofugal projections during early cortical 7. Rathelot JA, Strick PL (2009) Subdivisions of primary motor cortex based on cortico- τ Eur J Neurosci – motoneuronal cells. Proc Natl Acad Sci USA 106:918–923. development in a -GFP-transgenic mouse. 25:17 30. 8. O’Leary DD, Koester SE (1993) Development of projection neuron types, axon 22. Bulfone A, et al. (1998) An olfactory sensory map develops in the absence of normal pathways, and patterned connections of the mammalian cortex. Neuron 10:991–1006. projection neurons or GABAergic interneurons. Neuron 21:1273–1282. 9. McConnell SK (1995) Constructing the cerebral cortex: Neurogenesis and fate 23. Iwasato T, et al. (2004) Dorsal telencephalon-specific expression of Cre recombinase in determination. Neuron 15:761–768. PAC transgenic mice. Genesis 38:130–138. 10. Rash BG, Grove EA (2006) Area and layer patterning in the developing cerebral 24. Kawamoto S, et al. (2000) A novel reporter mouse strain that expresses enhanced cortex. Curr Opin Neurobiol 16:25–34. green fluorescent protein upon Cre-mediated recombination. FEBS Lett 470:263–268. 11. Molyneaux BJ, Arlotta P, Menezes JR, Macklis JD (2007) Neuronal subtype 25. Catsman-Berrevoets CE, Kuypers HGJM (1981) A search for corticospinal collaterals to fi Nat Rev Neurosci – speci cation in the cerebral cortex. 8:427 437. thalamus and mesencephalon by means of multiple retrograde fluorescent tracers in 12. Molyneaux BJ, Arlotta P, Hirata T, Hibi M, Macklis JD (2005) Fezl is required for the cat and rat. Brain Res 218:15–33. fi Neuron – birth and speci cation of corticospinal motor neurons. 47:817 831. 26. Johnson MB, et al. (2009) Functional and evolutionary insights into human brain 13. Chen JG, Rasin MR, Kwan KY, Sestan N (2005) Zfp312 is required for subcortical development through global transcriptome analysis. Neuron 62:494–509. axonal projections and dendritic morphology of deep-layer pyramidal neurons of the 27. Donoghue MJ, Rakic P (1999) Molecular evidence for the early specification of cerebral cortex. Proc Natl Acad Sci USA 102:17792–17797. presumptive functional domains in the embryonic primate cerebral cortex. J Neurosci 14. Chen B, Schaevitz LR, McConnell SK (2005) Fezl regulates the differentiation and axon – targeting of layer 5 subcortical projection neurons in cerebral cortex. Proc Natl Acad 19:5967 5979. Sci USA 102:17184–17189. 28. Gong S, et al. (2003) A gene expression atlas of the central nervous system based on 15. Kwan KY, et al. (2008) SOX5 postmitotically regulates migration, postmigratory bacterial artificial chromosomes. Nature 425:917–925. differentiation, and projections of subplate and deep-layer neocortical neurons. Proc 29. McKenna WL, et al. (2011) Tbr1 and Fezf2 regulate alternate corticofugal neuronal Natl Acad Sci USA 105:16021–16026. identities during neocortical development. J Neurosci 31:549–564.

3046 | www.pnas.org/cgi/doi/10.1073/pnas.1016723108 Han et al. Supporting Information

Han et al. 10.1073/pnas.1016723108 SI Materials and Methods mRNA-Seq and Data Analysis. Total RNA was isolated from freshly −/− +/+ Immunostaining. Brains dissected from embryonic and neonatal dissected P0 neocortices of Tbr1 and Tbr1 littermates using animals were fixed by immersion in 4% paraformadehyde overnight a Qiagen RNeasy Mini Kit. Libraries were prepared using an at 4 °C. Adult brains were fixed by perfusion, dissected, and post- Illumina mRNA-Seq Sample Prep Kit. Amplified cDNA was fixed by immersion in 4% paraformadehyde overnight at 4 °C. size-selected at 250 bp and validated using the Agilent Bio- fi Brains were vibratome-sectioned at 70 μm (Leica VT1000S). The analyzer DNA 1000 system. The nal product was subjected to sections were blocked and immunostained. The following primary cluster generation using an Illumina Standard Cluster Genera- antibodies were used at the indicated dilution: anti-CldU (rat, tion Kit v4. Libraries were sequenced as single end 74mers using 1:100; Accurate Chemical), anti-GFP (chicken, 1:3,000; Abcam), the Illumina Genome Analyzer pipeline, and image analysis anti-SOX5 (rabbit, 1:100; Genway), anti-TBR1 (rabbit, 1:250; (Firecrest module), base-calling (Bustard module), and primary Abcam and Santa Cruz Biotechnology), anti-ZFPM2 (rabbit, sequence analysis (Gerald module) were performed. Reads were 1:250; Santa Cruz Biotechnology), anti-RFP (rabbit, 1:1,000; US mapped to the mouse reference genome (release mm9) with ELAND software. Filtered reads that were uniquely mapped to Biological), anti-IdU (mouse, 1:200; BD Biosciences), anti-CUX1 exons of Ensembl gene model with up to two mismatches were (rabbit, 1:150; Santa Cruz Biotechnology), anti-SATB2 (mouse, used for the quantification of gene expression. To detect dif- 1:200; Santa Cruz Biotechnology), anti-NF1B (rabbit, 1:300; Active − − ferentially expressed genes between Tbr1 / and Tbr1+/+,we Motif), and anti-BC11B (rat, 1:250; Santa Cruz Biotechnology). designed a three-step stringent filtering process. First, counts of ’ Plasmid Constructs. For TBR1 expression, full-length mouse Tbr1 mapped reads within genes were compared using Fisher s exact test (adjusted P ≤ 0.05). Second, reads per kb of exon model per cDNA (BC052737) was inserted into pCAGEN (1). For ChIP, 9C NL C a N-terminal V5 tag (GKPIPNPLLGLDST) was added to create million reads (RPKM = 10 / , where is the number of mappable reads that fall onto the gene’s exons, N is the total the pCAG-V5-Tbr1 construct. Luciferase reporter plasmids were number of mappable reads in the lane, and L is the sum of the made by inserting annealed 57-bp-long complementary oligos exons in base pairs) were used to quantify gene expression, and representing the TBR1 biding site near Fezf2 into the SalI site of the fold change was set as another filtering scale (|log2(fold pGL3-CMV plasmid. pGL3-CMV was made by replacing the change)| ≥0.5) (3). In our series of relative experiments, the SV40 promoter of pGL3-promoter plasmid (Promega) with a genes were finally identified as differentially expressed unless CMV promoter. they were shared by all groups. Layer Distribution Analysis. To quantify the distribution of neurons, n – mRNA Sequencing Uniformity. One gene with exons was assigned the P0 P14 neocortex was divided radially into 10 equal-sized as one vector, G = {R1,R2,...,Ri,...,Rn}, where Ri is the ratio bins from the pia to the upper edge of the white matter. For of the ith exon expression (in RPKM) relative to the mean ex- E15.5 analyses, the pia to the upper edge of the intermediate pression of n exons. The Ri values were sorted according to the zone was divided radially into six bins. The cells in each bin were exon order from 5′ to 3′. Isoforms were combined to obtain the fi quanti ed and reported as the percentage of total cells counted. longest gene. Because different genes generally had different values of n, they were zoomed into the same length by uniformly Thymidine Analog Labeling. Timed-pregnant mice were injected filling some gaps in genes with fewer exons (e.g., G = {R1,GAP, intraperitoneally with 40 mg/kg body weight of CldU. Brain sec- R2,...,GAP,Ri,...,GAP,Rn}). The uniformity of the G vector tions were subjected to acidic antigen retrieval before blocking. from 5′ to 3′, quantified by mean and standard variation, directly reflected the degree of mRNA sequencing uniformity. Retrograde Axonal Tracing. For in vivo retrograde tracing, rhodamine-conjugated LMS (Lumafluor) were injected into the ChIP-Seq and Data Analysis. Cell Preparation. N2A cells were trans- target region of neonatal or P3 live mice under ice anesthesia. The fected with pCAG-V5-Tbr1 plasmid using Lipofectamine 2000 brains of injected neonatal mice were dissected after survival for (Invitrogen). At 48 h after transfection, cells were harvested – fi fi 18 24 h and xed. The injected P3 mice were xed by perfusion using a silicon scraper, transferred to a 50-mL conical tube, fi at P7. The injection site of all injected brains was veri ed both crosslinked by adding 1/10 volume of crosslinking solution for 15 externally and in sections through the injection site. Only brains min, and quenched by adding 1/10 volume of a 1.25 M glycine fi that were clearly and speci cally injected at the correct location solution. Crosslinked cells were washed, flash-frozen in liquid were used. Quantitative analyses of retrogradely labeled neurons nitrogen, and stored at −80 °C. were consistently performed in the same area in the somato- ChIP. ChIP was performed following a protocol modified from Lee sensory cortex to allow direct comparisons of brains. Confocal et al. (4). Crosslinked cells were thawed on ice and lysed. Lysates images with an optical thickness of 2 μm were used for all were sonicated (with a Misonix 4000) to shear the chromatin to quantifications. A square box encompassing the entire cortical 200∼500 bp for ChIP-Seq. For each ChIP reaction, 1000 μgof plate from the upper edge of the white matter to the upper edge sheared chromatin was incubated overnight at 4 °C with Dynal of the marginal zone was overlaid, and all neurons within the box Protein A/G beads bound to anti-V5 antibody (rabbit, 1:10; Ab- were quantified. cam) to precipitate V5-TBR1. Dynabeads were washed extensively and then eluted at 65 °C for 15 min with occasional vortexing. Luciferase Assays. N2A cells were plated and transfected with Eluted protein–chromatin complexes were reverse-crosslinked firefly luciferase (pGL3-CMV) containing a TBR1 binding site as by overnight incubation at 65 °C. ChIPed and input DNA were described previously (2). Transfected cells were lysed and as- treated with RnaseA and proteinase K and then purified. sayed 24 h later using a Promega Luciferase Reporter Kit. qRT- Library Preparation and Sequencing. An Illumina ChIP-Seq library PCR of pGL3-CMV was performed to normalize for trans- preparation kit was used in accordance with the manufacturer’s fection efficiency. instructions with modifications. DNA fragments were end-

Han et al. www.pnas.org/cgi/content/short/1016723108 1of7 repaired, “A” bases were added to the 3′ end, and sequencing ChIP qRT-PCR. Cell Preparation. Primary cortical neurons were pre- adaptors were ligated. To reduce the free adaptor dimers that pared from microdissected E13.5 Fezf2-Gfp transgenic cortices could be preferentially amplified during PCR, the amount of and transfected with pCAG-V5-Tbr1 plasmid DNA using the adaptors added was reduced to 1:30 or 1:40 instead of 1:10. After amaxa Mouse Neuron Nucleofector Kit (VPG-1001; Lonza). At ligation, PCR amplification was performed first, and 250- to 350- 48 h after transfection, cells were collected as described above. bp fragments were size-selected. Gel extraction was performed at ChIP. Primary neuronal lysates were sonicated to shear the chro- room temperature to avoid the decreased representation of an matin to 200-1,000 bp for ChIP–qRT-PCR. ChIP was performed as A+T-rich sequence. Libraries were sequenced with an Illumina described above for ChIP-Seq. GAII Sequencing System in accordance with the manufacturer’s qRT-PCR of ChIPed DNA. For the qRT-PCR reactions, 500 pg of ChIP protocol. samplewasamplifiedusingSYBRgreenPCRMasterMix(Roche). Data Analysis. All 36-bp reads were aligned to the mouse reference Thermocycling was carried out using the Applied Biosystems 7900 genome (NCBI build 37, mm9) using the bowtie program. Only System. Input DNA values were used for normalization. uniquely aligned reads were used for subsequent analysis. The Nuclear Size Analysis. Coronal sections from LMS-traced Tbr1+/+; − − numbers of reads that fell into each 50-bp shift window were Golli-Gfp and Tbr1 / ;Golli-Gfp littermate brains (n = 2 for each calculated across whole chromosomes. A single bedGraph data genotype) were immunostained for GFP and ZFPM2 and nu- file was created for each reaction and uploaded to the UCSC + clear-stained using DAPI. WT L5 CS neurons (LMS ), WT L6 Genome Browser (http://genome.ucsc.edu/). Golli-Gfp+ + Tbr1−/− Determination of TBR1 Consensus Binding Sequence. neurons ( and ZFPM2 ), and L6 CS neurons All 36-bp reads + Golli-Gfp+ + from V5 antibody ChIP were aligned to the mouse genome mm9 (LMS , , and ZFPM2 ), were examined by confocal n as described above. All reads for input were aligned as well. Then microscopy ( = 30 cells per cell type). The plane of focus was a sliding-window approach, with a window size of 200 bp and adjusted precisely so that the nucleus of the cell of interest was at a shift step of 50 bp, was used to detect enriched genome regions. its roundest and largest. The length of the longest diameter Windows enriched in the V5 experiment by 20-fold greater than drawn on the nucleus was used for analysis. expected were selected. Overlapping enriched windows were Generation of Fezf2 Conditional Knockout Mice. The targeting vector merged into one enriched region. To minimize background sig- for conditional inactivation of Fezf2 was constructed by a re- nals, regions enriched in the control experiment by fivefold combineering-based method (5) with a bacterial artificial chro- greater than expected were removed from the enriched list. A mosome (BAC; clone RP23-141E17). Exon 2, which encodes proximal gene with minimum distance was selected for each amino acids 1–279 of 455, was targeted for removal. An 11-kb enriched region. Genomic information of all genes was down- fragment containing exon 2 was retrieved from the BAC clone loaded from the UCSC Genome Browser. and inserted into the PL253 vector by recombination in the fi To speci cally select for sites to which TRB1 binding led to bacterial strain SW102. PGK-neo was used as a positive selection changes in gene expression, only genes that were proximal to marker, and the MC1-TK (thymidine kinase) cassette was used enriched regions were chosen for determination of the consensus as a negative selection marker. The loxP/Frt-flanked positive binding sequence. Only genes with differential expression be- − − selectable marker and the loxP site for conditional deletion of tween Tbr1+/+1 and Tbr1 / 1 and also between Tbr1+/+2 and −/− exon 2 were inserted as described in Fig. 8A. The linearized Tbr1 2 were selected for the final list of candidate genes regulated by Tbr1. All enriched regions whose proximal genes were targeting construct was electroporated into embryonic stem cells, in the foregoing candidate list were selected. The PRIORITY which were then subjected to G418 selection. Clones positive for fi program (http://www.cs.duke.edu/∼amink/software/priority/) was homologous recombination were identi ed by long-distance used to search for consensus sequences, with the following pa- PCR and confirmed by sequencing before blastocyst injection. rameters: a motif length of 20, a flat prior, a third-order back- Cortex-specific deletion of exon 2 was achieved using the Emx1- ground model, 50 trials, and 10,000 iterations per trial. For the Cre PAC transgene. TBR1 binding site, the multispecies conservation analysis was qRT-PCR. Total RNA isolated for mRNA-Seq was also used for performed using the 30-way Multiz Alignment and Conservation qRT-PCR. Template cDNA was synthesized, and predesigned track in the UCSC Genome Browser. primer and probe sets were used.

1. Matsuda T, Cepko CL (2007) Controlled expression of transgenes introduced by in vivo 6. Kispert A, Hermann BG (1993) The Brachyury gene encodes a novel DNA binding electroporation. Proc Natl Acad Sci USA 104:1027–1032. protein. EMBO J 12:4898–4899. 2. Abelson JF, et al. (2005) Sequence variants in SLITRK1 are associated with Tourette’s 7. Hevner RF, et al. (2001) Tbr1 regulates differentiation of the preplate and layer 6. syndrome. Science 310:317–320. Neuron 29:353–366. 3. Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B (2008) Mapping and 8. Chen B, Schaevitz LR, McConnell SK (2005) Fezl regulates the differentiation and axon quantifying mammalian transcriptomes by RNA-Seq. Nat Methods 5:621–628. targeting of layer 5 subcortical projection neurons in cerebral cortex. Proc Natl Acad Sci 4. Lee TI, et al. (2006) Control of developmental regulators by Polycomb in human USA 102:17184–17189. embryonic stem cells. Cell 125:301–313. 9. Molyneaux BJ, Arlotta P, Hirata T, Hibi M, Macklis JD (2005) Fezl is required for the 5. Liu P, Jenkins NA, Copeland NG (2003) A highly efﬁcient recombineering-based birth and speciﬁcation of corticospinal motor neurons. Neuron 47:817–831. method for generating conditional knockout mutations. Genome Res 13:476–484.

Han et al. www.pnas.org/cgi/content/short/1016723108 2of7 − − Fig. S1. SOX5 protein levels were largely unaltered in the late embryonic and neonatal Tbr1 / neocortex. (A and B) Immunoﬂuorescent staining for SOX5 at E16.5 revealed intense nuclei staining in the cortical plate of Tbr1−/−, comparable to that seen in Tbr1+/+ embryos. (C) Neocortical lysates were collected from Tbr1−/− and Tbr1+/+ P0 littermates. Western blot analysis revealed that SOX5 protein levels were not signiﬁcantly altered in Tbr1−/− at P0. GAPDH was used as a loading control. (Scale bars: 100 μm.)

Fig. S2. Direct TBR1 binding of a novel 10-bp consensus binding site near Fezf2 was deﬁned by global analysis of ChIP-Seq reads. (A) A computational screen for an enriched consensus motif with nucleotide bias using Weblogo 3. This novel TBR1 binding sequence motif is similar to the reported TBR1 binding site in Xenopus determined using random oligonucleotides (6). (B) A region 4.5 kb downstream of the Fezf2 TSS with signiﬁcant enrichment of V5-TBR1 ChIP-Seq reads over input DNA. This region contains a consensus TBR1 binding sequence motif (red). (C) Sequence alignment of the TBR1 binding motif (red box) across 21 mammalian species demonstrating high conservation.

Han et al. www.pnas.org/cgi/content/short/1016723108 3of7 Fig. S3. TBR1 was expressed in a pattern highly complementary to Fezf2 and was excluded from CS neurons in early postnatal neocortex. (A) Neocortices of Fezf2-Gfp (green) transgenic mice were immunostained for TBR1 (red). At E15.5, before the onset of postmigratory down-regulation of Fezf2 in L6, 63.5% of TBR1+ L6 neurons coexpressed high levels of Fezf2-Gfp (yellow). At P0, Fezf2-Gfp was highly expressed in L5 neurons but was down-regulated in L6 and SP neurons, which expressed TBR1 abundantly. Only 7.7% of TBR1+ neurons coexpressed Fezf2-Gfp at this stage. This almost mutually exclusive pattern of gene expression in L5 and L6 was maintained into the early postnatal stages. Starting at P7, TBR1 was also expressed at low levels in L2–L4, in which Fezf-Gfp was completely absent. (B) The retrograde tracer LMS (red) was injected into the cervical spinal cord of live mice at P3, and the mice were killed at P7. All retrogradely traced CS neurons were positioned in L5b; none expressed detectable TBR1 (green). This absolute exclusion of TBR1 from early postnatal CS neurons and its highly complementary expression with Fezf2 suggest that TBR1 is a compelling candidate for putatively restricting high Fezf2 expression to L5 neurons. (Scale bars: 100 μm.)

− − Fig. S4. Thymidine analog birth-dating conﬁrmed migration defects and revealed an ectopic band of cells containing early-born neurons in Tbr1 / neocortex (Nctx). (A and B) IdU was injected into timed-pregnant females at E11.5, when SP and L6 neurons are generated. (A) At P0, neurons born at E11.5 were positioned in the SP and the lower portion of L6 in the Tbr1+/+.(B)InP0Tbr1−/−, neurons born at E11.5 were positioned primarily in a thin ectopic band of cells (red brackets) located in the middle of the cortical plate, indicating defects in neuronal migration and preplate segregation. In addition to this band of cells, a considerable proportion of early-born neurons were dispersed to all cortical layers. (C and D) CldU was injected into timed-pregnant females at E15.5, when upper layer − − neurons are born. In P0 Tbr1+/+, these late-born neurons migrated to L2–L3 (C). In P0 Tbr1 / , most of these late-born neurons were ectopically positioned at deep levels of the neocortex (D). Although their distribution was more dispersed, a small proportion of E15.5-labeled neurons migrated to L2. (Scale bar: 500 μm.)

Han et al. www.pnas.org/cgi/content/short/1016723108 4of7 Fig. S5. ZFPM2 (also known as FOG2) is a highly speciﬁc marker of corticothalamic projection neurons. Neurons of distinct projectional subclasses in P7 WT neocortex were retrogradely labeled using the tracer LMS (red). LMS was injected into the contralateral neocortex, ipsilateral dorsal thalamus, or cervical spinal cord at P3 to label corticocortical, corticothalamic, or CS neurons, respectively. Coronal sections from traced brains were immunostained for ZFPM2 (green). None of the retrogradely labeled corticocortical or CS neurons was positive for ZFPM2 (A and C). In contrast, virtually all neurons labeled by LMS injected into the dorsal thalamus were ZFPM2+ (B, solid arrowheads), indicating that ZFPM2 is a highly speciﬁc marker of corticothalamic projection neurons. (Scale bar: 50 μm.)

− − − − Fig. S6. Expression of layer-speciﬁc markers in Tbr1 / neocortex. P0 neocortices were immunostained with the indicated primary antibodies. In Tbr1 / neocortex, most deep-layer neurons were positioned within clusters, as described previously (7). The molecular markers of L6 neurons were differentially affected. Expression of the L6 markers FOXP2 and NFIB was diminished, whereas that of ZFPM2 was normal. Expression of the L5 marker BCL11B was slightly increased. Most upper-layer neurons of the Tbr1−/− neocortex (denoted by CUX1 and SATB2) did not migrate to their normal destination; instead, they were dispersed, mostly located in the deeper portion of the neocortex and largely absent from the clusters containing L5 and L6 neurons. The expression levels of CUX1 and SATB2 were not diminished, however. (Scale bar: 50 μm.)

Han et al. www.pnas.org/cgi/content/short/1016723108 5of7 Fig. S7. Speciﬁc defects in Tbr1−/− subcortical axon tracts. P0 coronal (A) and sagittal (B and C) sections of Tbr1−/− and Tbr1+/+ littermates transgenic for Fezf2- − − Gfp.(A–D)IntheTbr1+/+ brain, Fezf2-Gfp+ axons abundantly innervated the thalamus (solid arrowheads). In the Tbr1 / brain, no Fezf2-Gfp+ axons innervated the dorsal thalamus, indicating complete loss of the corticothalamic tract. Ectopic subcortical axons innervated the basal forebrain (arrows), and the ste- − − reotypical organization of the internal capsule was lost. (E and F) The CS tract of the Tbr1 / (open arrowheads) was similar to that of the Tbr1+/+ (solid ar- − − rowheads). Some of the Tbr1 / CS axons were less tightly bundled and followed a more dorsal trajectory through the ventral pons. (G and H) The thickness of the CS tract was unaffected in Tbr1−/−, indicating that the number of CS axons was not signiﬁcantly altered. (Scale bars: 500 μm.)

Fig. S8. Cortex-specific conditional deletion of Fezf2 phenocopies germ-line Fezf2 null mutation. (A) Schematic of the targeting strategy used for generating a floxed Fezf2 allele. Cortex-specific deletion of Fezf2 was achieved using the Emx1-Cre transgene. (B) Correct insertion of loxP sites and presence of the Emx1- Cre transgene were confirmed by PCR of genomic DNA extracted from tail biopsy specimens. CRE-mediated recombination exon 2 in the cortex was confirmed by PCR of DNA extracted from neocortices of Fezf2flox/flox;Emx1-Cre mice. (C–E) To examine the pancortical pattern of axonal connectivity, Fezf2flox/flox;Emx1- Cre mice were further crossed with CAG-Cat-Gfp transgenic mice. CRE-mediated recombination occurred in all neocortical projection neurons and activated GFP expression from the CAG-Cat-Gfp transgene (C). Analysis of GFP-filled axons of cortical projection neurons revealed that the CS tract was completely absent (D and E), phenocopying germ-line knockouts of Fezf2 (8, 9). (Scale bars: 200 μm.)

Han et al. www.pnas.org/cgi/content/short/1016723108 6of7 − − Fig. S9. Schematic summary of Tbr1 / defects in regulation of Fezf2 transcription (A) and the laminar origin of distinct subcortical projection subtypes (B). In L6 neurons of the Tbr1+/+ neocortex, TBR1 binds to an evolutionarily conserved consensus motif near the Fezf2 locus and represses the expression of Fezf2.In Tbr1−/− neocortices, L6 neurons aberrantly turn on high Fezf2 expression. Concomitant with increased Fezf2 expression, Tbr1−/− L6 neurons project ectopically to the hypothalamus (Hyp) and brainstem.

Han et al. www.pnas.org/cgi/content/short/1016723108 7of7